| Message ID | 1554433498-58551-1-git-send-email-guohanjun@huawei.com (mailing list archive) |
|---|---|
| State | RFC |
| Headers | show |
| Series | [RFC] arm32: spectre-v2: using raw_smp_processor_id() for preemptible | expand |
diff --git a/arch/arm/include/asm/system_misc.h b/arch/arm/include/asm/system_misc.h index 66f6a3a..55e3b2b 100644 --- a/arch/arm/include/asm/system_misc.h +++ b/arch/arm/include/asm/system_misc.h @@ -22,7 +22,7 @@ static inline void harden_branch_predictor(void) { harden_branch_predictor_fn_t fn = per_cpu(harden_branch_predictor_fn, - smp_processor_id()); + raw_smp_processor_id()); if (fn) fn(); }
When enabling CONFIG_DEBUG_PREEMPT, there is a BUG report: BUG: using smp_processor_id() in preemptible [00000000] code: remap_file_page/29696 caller is __do_user_fault+0x34/0x114 It turns out that in harden_branch_predictor() using smp_processor_id() to get current cpu's callback for harden_branch_predictor(). This seems to be a false positive report as we need this running cpu's callback to clean the cache or BTB event it's preemptible, so using raw_smp_processor_id() instead to eliminate the false positive report. Signed-off-by: Hanjun Guo <guohanjun@huawei.com> --- Hi, Not sure I'm doing the right thing, so send as RFC. arch/arm/include/asm/system_misc.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)